Simulation of key performance characteristics under uncertainty

Abstract

The engineering designer frequently has to deal with multiple uncertainties, especially at the early design stages or when working on products for which the body of knowledge is incomplete. Wash performance of dishwashers is one such case. This paper demonstrates the application of the design for system integrity (DSI) methodology to create a model to help to predict wash performance. This is a complex problem because of the diverse uncertainties: stochastic, epistemic, and abstraction. The DSI model embodied the belief that two principal mechanisms contribute to wash performance: the effectiveness of soil removal, and the effectiveness of the rinse process (i.e. lack of re-deposition of soil). The model was calibrated against wash data from a known machine and then predicted the wash performance of a different brand of machine with qualitative differences in geometry. The results show that it is possible to develop models for highly uncertain systems, and this has important implications for early engineering design among other domains. Being able to simulate performance under considerable uncertainty, even if imperfectly, is potentially an important tool for the early design stages as it could indicate, firstly, promising design avenues and, secondly, the risk in the design and the degree to which design issues have been resolved.